Establishment of the In Vivo Efficacy of Pretargeted Radioimmunotherapy Utilizing Inverse Electron Demand Diels-Alder Click Chemistry

This article is featured in Highlights of This Issue, p. 1

For over three decades, the remarkable selectivity and affinity of antibodies for tumor biomarkers have made them attractive vectors for the selective delivery of therapeutic radiation to cancer cells (1, 2). A wide variety of antibodies have been labeled with a range of therapeutic isotopes. Indeed, a number of therapeutic radioimmunoconjugates are currently the subject of clinical trials, and two constructs—⁹⁰Y-ibritumomab tiuxetan and ¹³¹I-tositumomab—have been approved for the treatment of non-Hodgkin lymphoma (3, 4). However, an important limitation compromising targeted radioimmunotherapy (RIT) is that immunoglobulins can take multiple days to reach their optimal biodistribution in vivo. This relatively slow pharmacokinetic profile mandates the use of therapeutic isotopes with long physical half-lives, such as ⁹⁰Y (t_1/2 ∼ 2.7 days), ¹⁷⁷Lu (t_1/2 ∼ 6.6 days), and ¹³¹I (t_1/2 ∼ 8.0 days). This combination of slow pharmacokinetics, long physical half-lives, and therapeutic radiation can potentially result in prohibitively high radiation doses to healthy organs, particularly bone marrow (5). An effective radioimmunoconjugate must therefore strike a balance between delivering a therapeutic dose of radiation to the tumor while avoiding, or at least minimizing, radiation-related side effects. This issue becomes particularly important in the context of solid lesions, for which effective RIT has long proven elusive due to slow uptake, radiation resistance, and poor tumor penetration (6, 7). In principle, these obstacles could be overcome by injecting higher activities; however, the suboptimal tumor-to-tissue activity concentration ratios of traditional radioimmunoconjugates often preclude this course, as dose-limiting radiation toxicities would be reached before effective therapy could be achieved.

In response to these issues, considerable attention has been dedicated to the creation of RIT strategies that reduce the radiation burden to healthy tissues, most notably the development of radiolabeled antibody fragments (8, 9). Another promising approach is pretargeted radioimmunotherapy (PRIT; ref.10). The aim of PRIT is to harness the tumor-targeting properties of antibodies while avoiding their pharmacokinetic drawbacks by decoupling the targeting vector from the radioisotope. More specifically, PRIT strategies typically employ four steps (Fig. 1A): (i) the administration of an unlabeled antibody with the ability to bind both an antigen and a radioligand; (ii) the slow accumulation of the antibody in the tumor and its concomitant clearance from the blood; (iii) the subsequent administration of a small-molecule radioligand; and (iv) the rapid binding of the radioligand to the antibody at the tumor site and the rapid excretion of any excess radioligand. Administration of the radioisotope in the form of a small molecule with rapid pharmacokinetics rather than a large molecule (i.e., the antibody) with slow kinetics is the key feature of this paradigm. As a result, PRIT yields high activity concentrations in tumors while keeping radiation doses to healthy organs low. Until recently, three approaches to in vivo pretargeting have dominated the field: bispecific antibodies engineered to bind both an antigen and a radiolabeled hapten, streptavidin-labeled antibodies with biotin-based radioligands, and oligonucleotide-bearing antibodies with radioligands modified with complementary sequences. All three strategies have proven effective in preclinical studies, yet each has been limited somewhat by various obstacles, such as the immunogenicity of streptavidin-based immunoconjugates (11–15).

The past 5 years have witnessed the emergence of a new approach to in vivo pretargeting based on the inverse electron demand Diels-Alder (IEDDA) reaction between tetrazine (Tz) and trans-cyclooctene (TCO; Fig. 1B; refs.16, 17). This straightforward and modular methodology is predicated on the extraordinary selectivity and rapidity (k₁ > 30,000 M⁻¹s⁻¹) of the bioorthgonal cycloaddition between a TCO-modified antibody and a tetrazine-bearing radioligand (Fig. 1A; refs.18–21). The earliest reports of this methodology focused on pretargeted SPECT imaging using an ¹¹¹In-labeled radioligand and a TAG72-targeting CC49-TCO immunoconjugate (22). More recently, our laboratories have developed highly effective pretargeted PET imaging strategies for both colorectal carcinoma and pancreatic ductal adenocarcinoma (PDAC) using ⁶⁴Cu- and ¹⁸F-labeled tetrazine radioligands (23–25). In both cases, the pretargeting approaches identify tumor tissue extremely well, yielding images with high tumor-to-background contrast at only a fraction of the radiation dose produced by directly labeled radioimmunoconjugates. Given the success of these pretargeted imaging modalities, the logical next step is to leverage this technology for the development of a safe and effective approach to PRIT (26). Notably, although Rossin and colleagues have successfully demonstrated the feasibility of ¹⁷⁷Lu-based PRIT utilizing this methodology, to the best our knowledge no in vivo therapy studies have yet been published (27, 28).

In this article, we present the development, optimization, and in vivo validation of a pretargeted approach for the RIT of PDAC based on bioorthogonal click chemistry. We have used ¹⁷⁷Lu-labeled tetrazine radioligands (Fig. 1C) and a TCO-modified immunoconjugate of 5B1, a fully human antibody that targets the PDAC biomarker carbohydrate antigen 19.9 (CA19.9; refs.23, 24, 29–31). We demonstrate that this approach elicits a dose-dependent therapeutic response in mice bearing CA19.9-expressing BxPC3 human PDAC xenografts.

All studies were performed in accordance with Memorial Sloan Kettering Institutional Animal Care and Use Committee. All in vivo studies were carried out in female, athymic nude mice [Crl:NU(NCr)-Foxn1^nu, 6–8 weeks) bearing subcutaneous BxPC3 xenografts that were inoculated on the right flank, as previously described (29, 31). BxPC3 cells obtained from ATCC were obtained in November 2014, validated via STR analysis, and used within 4 weeks of resuscitation. BxPC3 cells were grown in RPMI medium modified to contain 4.5 g/L glucose, 1.5g/L sodium bicarbonate and supplemented with 10% (v/v) heat-inactivated FCS, 100 IU penicillin, 100 μg/mL streptomycin, 10 mmol/L HEPES, and 10 cc/L nonessential amino acids. Mice were xenografted subcutaneously with 5 × 10⁶ cells, suspended in 150 μL of a solution containing a 1:1 mixture of Matrigel (Becton Dickinson), and cell culture medium. BxPC3 tumors were grown for 21 to 28 days postimplantation before imaging or biodistribution.

The N-hydroxysuccinyl ester of TCO [(E)-cyclooct-4-enyl 2,5-dioxo-1-pyrrolidinyl carbonate; TCO-NHS)] was conjugated to 5B1 as previously described (23). The average number of TCO molecules per antibody was estimated by incubating the 5B1-TCO with >150-fold excess of a Tz-functionalized Alexa Fluor 680 (Tz-AF680) and quantifying the ratio of Tz-AF680 per antibody, after purification by gel filtration, via UV-Vis as previously described (23–25). Further details are provided in the Supporting Information.

Both DOTA-PEG₇-Tz and CHX-A″-DTPA-PEG₇-Tz were prepared via a three-step synthesis that is fully described in Supplementary Fig. S1. In short, a commercially available amine-reactive N-hydroxysuccinyl tetrazine (Tz-NHS) was coupled to a mono-Boc protected, bis-amino PEG₇ molecule to form Tz-PEG₇-NHBoc (1) which was subsequently deprotected in a mixture of methylene chloride and trifluoroacetic acid to form a common intermediate, Tz-PEG₇-NH₂ (2). Then, the p-isothiocyanate of DOTA or CHX-A″-DTPA was coupled to yield the products; DOTA-PEG₇-Tz and CHX-A″-DTPA-PEG₇-Tz. All products and intermediates were characterized by ¹H-NMR (Supplementary Fig. S2), ¹³C-NMR, ESI-MS, and high-resolution mass spectrometry (HR-MS).

DOTA-PEG₇-Tz and CHX-A″-DTPA-Tz were radiolabeled (Supplementary Fig. S3) by diluting stock solution (10 mg/mL in DMSO) into ammonia acetate buffer (200 mmol/L, pH 5.4, 100 μL) prior to addition of ¹⁷⁷LuCl₃ in a 0.05 mol/L HCl solution (0.5–5 μL). The labeling solution was incubated at room temperature for 20 minutes before quenching with EDTA (50 mmol/L, pH 5.5, 10 μL). The products were isolated via HPLC (0–100% MeCN, 15 minutes), and the solvent was removed on a rotary evaporator before reconstituting in 0.9% saline for injection.

The partition coefficient (log D) for both radioligands was determined in PBS and 1-octanol. In short, a small amount of radioligand (∼100 μCi) was added to a 1:1 mixture of PBS and 1-octanol and the sample was vortexed for 10 minutes, and the layers separated by centrifugation (1,000 rpm for 10 minutes). The layers were then placed into separate tubes for quantification by gamma counting. In addition, the in vitro stability of each radioligand was analyzed in both PBS (pH 7.4) and human serum at 37°C up to 48 hours (Supplementary Table S1). Aliquots of radiolabeled ligands (1.0 mCi) were added to PBS or serum and incubated at 37°C. At various time points an aliquot was removed and analyzed by radio-iTLC. All experiments were performed in triplicate; additional details may be found in the Supplementary Information.

Tumor-bearing mice were injected via the lateral tail vein with 5B1-TCO (100 or 200 μg). For all biodistribution studies, mice that received 5B1-TCO were administered 40–50 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz or ¹⁷⁷Lu-CHX-A″-DTPA-Tz in an approximately 1.5:1 ratio of radioligand to 5B1-TCO (either 1 or 2 nmol), which was delivered 72 hours after administration of the 5B1-TCO (Fig. 2; Supplementary Fig. SI4–SI6; and Supplementary Table S2). In addition, one cohort was administered 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz to ensure that increasing the specific activity did not cause a deviation in total the biodistribution (this method was established as the optimal dosing strategy based on the in vivo behavior of 5B1 in previous studies; refs.23, 24). Tissues were collected and analyzed as previously described between 4 and 120 hours. Further details are outlined in the supporting information.

The biodistribution data for the ¹⁷⁷Lu-Tz-PEG₇-DOTA–based pretargeting system were expressed as normal-organ mean standard uptake values (SUV) versus time postadministration. Assuming, in first order, that SUVs are independent of body mass and thus the same among species, the mean SUV in mouse organ i, SUV_i, was converted to the fraction of the injected dose in the corresponding human organ I, FID_I, using the organ and total-body masses of the 70-kg Standard Man anatomic model and the following formula:

These data (corrected for radioactive decay) were fit to exponential time–activity functions. The cumulated activity, or residence time, in human organ i (t_i, in μCi-h/μCi) was calculated by integrating the time–activity function in organ i, replacing the biological clearance constant, (λ_b)_x for each component, x, of the fitted exponential function with the corresponding effective clearance constant, (λ_e)_x [(λ_e)_x = (λ_b)_x + λ_p, where λ_p is the physical decay constant of the radionuclide]. The resulting organ residence times were entered into the OLINDA computer program to yield the mean organ absorbed doses and effective dose in rad/mCi and rem/mCi, respectively (Tables 1 and 2; ref.32).

Tumor volumes for all mice were determined via caliper measurements of the longest dimension (x) and shortest dimension (y) and using the equation tumor volume (mm³) = x/2 × y². After initial tumor measurements, mice were randomized into five cohorts (n = 7–8 per cohort), ensuring all cohort had approximately equal average tumor volumes. Seventy-two hours prior to administration of the radioligands, the three cohorts that were selected to receive therapy were administered 5B1-TCO (200 μg; 1.3 nmol), whereas the two control cohorts were administered vehicle only (0.9% sterile saline). Seventy-two hours later (day 1), the tumors dimensions were measured again, and those resulting volumes set as the initial tumor volumes to which the relative growth was compared throughout the study. Immediately after day 1 tumor measurements, mice in the therapy cohorts were then administered 400, 800, or 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz (1.92 ± 0.06 nmol, 1.44 ± 0.05 Tz:mAb ratio). The control cohorts that had not been previously administered 5B1-TCO were injected with either vehicle (0.9% sterile saline) or 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz (1.83 nmol). Tumor volume was monitored via caliper measurement every 3 to 7 days up to the predetermined endpoint of tumor volume (1,000 mm³) or 51 days (Fig. 3). All mice were assessed twice per week throughout the study for outward signs of toxicity, including lethargy, loss of appetite, and decreasing body weight. The endpoint for individual mice was defined as the point at which the volume of the xenograft was measured to be ≥1,000 mm³, and the study was terminated when 50% of both control groups had reached the predetermined endpoint.

⁸⁹Zr-DFO-5B1 (⁸⁹Zr-5B1) was prepared as previously described (29). For this study, mice that still harbored viable tumor tissue of approximately the same size (200–450 mm³) were selected, including one from each control group and two each from the cohorts that received 800 or 1,200 μCi of radioligand in the in vivo PRIT study. The selected mice were injected with ⁸⁹Zr-5B1 (118 ± 3 μCi, ∼35 μg, ∼0.23 nmol), and PET images were acquired at 120 hours postinjection as previously described (Fig. 4A-B and Supplementary Fig. S7; refs.29, 31). Further details regarding image acquisition, reconstruction, and processing are provided in the Supplementary Information.

BxPC3 xenografts were excised and one each was either embedded in Tissue–Plus OCT compound and frozen on dry ice or fixed with 4% paraformaldehyde (PFA) solution and subsequently sectioned (10 μm). Masson's trichrome staining, autoradiography, and anti-CA19.9 immunofluorescence staining were performed on sequential sections as previously described (Fig. 4C; refs.29, 31). Further details may be found in Supplementary Information.

To analyze the effect of therapy within each cohort, the therapeutic response was assessed using a random effect model (33). In this model it is assumed that the log volume–time profile of each mouse deviates in a random fashion from an average profile. The model for mouse i, in which μ is the average intercept, α_i is a random intercept (accounting for random effect) for each mouse, β is the slope (treatment effect over time), and c_i is a random error term, is

Using a Likelihood ratio test, the above model was used to test for a significant change in log tumor volume over time, within each cohort and for pair-wise comparisons between cohorts. For each cohort, we use the above model to estimate the doubling or halving time and the delta method to calculate a 95% confidence interval for our estimate (34). All models were fit using the lme4 package in R version 3.1.1.

For this investigation, 5B1-TCO was synthesized via conjugation of the antibody with TCO-NHS and characterized in vitro as previously reported (23).

Radioligand precursors were successfully synthesized in three steps: the coupling of Tz-NHS and O-(2-aminoethyl)-O'-[2-(bocamino)ethyl]hexaethylene glycol to form Tz-PEG₇-NHBoc, the removal of the protecting group with TFA/CH₂Cl₂, and the ligation of Tz-PEG₇-NH₂ with either p-NCS-Bn-DOTA or p-NCS-Bn-CHX-A″-DTPA (Supplementary Fig. S1). In this way, Tz-PEG₇-CHX-A″-DTPA (1) and Tz-PEG₇-DOTA (2) were synthesized in 87% and 73% yield, respectively. In both cases, the chemical structure of all compounds was confirmed via ¹H-NMR (Supplementary Fig. S2), ¹³C-NMR, ESI-MS, and HR-MS.

The chelator-modified tetrazines were radiolabeled via incubation with [¹⁷⁷Lu]-LuCl₃ for 10 minutes at 37°C in 250 mmol/L NH₄OAc, pH 5.5 (Supplementary Fig. S3). ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz (¹⁷⁷Lu-1) was produced in >99% radionuclidic purity, 99.5% radiochemical yield, and a specific activity of 383 mCi/μmol (n = 3), whereas ¹⁷⁷Lu-DOTA-PEG₇-Tz (¹⁷⁷Lu-2) was produced in >96% radionuclidic purity, 96.9 ± 0.74% radiochemical yield, and a specific activity of 466 ± 155 mCi/μmol (n = 3).

The partition coefficients of the radioligands were determined using 1-octanol and PBS (pH 7.4), yielding values for the two hydrophilic compounds of −3.7 ± 0.1 for ¹⁷⁷Lu-DOTA-PEG₇-Tz and −3.2 ± 0.2 for ¹⁷⁷Lu- CHX-A″-DTPA-PEG₇-Tz. These values suggest that there is little difference in the hydrophobicity of the ligands. Stability assays revealed that both radioligands are quite stable, with >85% and >80% of each radioligand remaining intact after an incubation of 48 hours in PBS and serum, respectively (Supplementary Table S1). These rates of decomposition are unlikely to hamper the in vivo performance of the system, as the clearance half-times of the radioligands are short. Ultimately, owing to the similar physical properties and the relative ease of synthesis and purification, we chose to evaluate both radioligands in preliminary in vivo investigations.

Biodistribution studies were performed with three goals in mind: (i) the identification of the appropriate dose of 5B1-TCO to give for PRIT in this xenograft model; (ii) a side-by-side comparison of both radioligands to determine the best candidate for in vivo PRIT studies; and (iii) the quantification of the uptake and clearance of the preferred radioligand for dosimetry calculations. We first performed a pilot study in which mice with BxPC3 xenograft-bearing mice were administered either 100 or 200 μg of 5B1-TCO, followed by 1.5 molar equivalents of ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz. We chose to analyze only one radioligand for this study to reduce the number of animals needed. We found that administration of 200 μg of 5B1-TCO led to significantly better tumor accumulation and similar or even better tumor-to-tissue activity concentration ratios (Supplementary Fig. S4A and S4B), consistent with previously reported PET imaging studies using 5B1-TCO as a targeting vector (23, 24).

Next, we compared the two radioligands using the optimal amount of 5B1-TCO. This investigation showed that pretargeting with ¹⁷⁷Lu-DOTA-PEG₇-Tz produced significantly higher tumor activity concentrations than pretargeting with ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz: 12.0 ± 5.3% ID/g and 4.3 ± 1.8% ID/g at 72-hour postinjection, respectively (Fig. 2; Supplementary Fig. S4C and S4D). This led us to choose the former for in vivo therapy studies. A final biodistribution was performed by administering 200 μg of 5B1-TCO followed 72 hours later by 1.5 molar equivalents of ¹⁷⁷Lu-DOTA-PEG₇-Tz and collecting tissues at five time points between 4 and 120 hours. The activity concentrations observed in the tumor tissue were high, reaching 16.8 ± 3.87% ID/g at 120 hours postinjection (Fig. 2; Supplementary Fig. S5; and Supplementary Table S2). Clearance from background tissue was rapid, with all tumor-to-tissue activity concentration ratios exceeding 15:1 at 120 hours (Fig. 2, inset).

The mean organ absorbed doses for mice and the extrapolation of this data to the 70 kg Standard Man model are reported in Tables 1 and 2, respectively. The ratio of the mean organ absorbed dose in mice relative to that to the tumor (“therapeutic index”) is also reported. Calculating the absorbed dose in humans using the dosimetry data suggests that a 150 mCi administration of radioligand would not exceed the MTD in off-target tissues in humans. Assuming a MTD for red marrow of 150 rad, the maximum tolerated administered activity would be approximately 700 mCi (Table 2; refs.7, 35).

To establish dose response using the optimized PRIT protocol, mice (n = 7–8 per cohort) were administered 5B1-TCO 72 hours prior to the injection of 400, 800, or 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz. Control groups were not administered 5B1-TCO and were instead injected either with vehicle or 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz. One mouse from the 800 μCi therapy cohort was euthanized after 24 days due to an infection, although it is unclear whether this was event was related to the therapy study. No other mice showed any outward signs of toxicity, and all mice maintained an acceptable body weight (>80% original weight) throughout the course of the study. Average weight loss for any cohort never exceeded 6%, and there was not any significant difference among any of the therapy or control cohorts at any time point. These results suggest that the treatment was well tolerated up to the highest administered therapeutic dose of 1,200 μCi.

Following the average tumor volume for the first 31 days after treatment clearly shows a differential response that improves (more dramatic growth delay or shrinkage) with increasing ¹⁷⁷Lu activity in the therapy cohorts (Fig. 3A and B). Analysis of the growth-delay effects across each cohort becomes convoluted after that point, as mice with tumors >1,000 mm³ are removed from that point on. As of 51 days posttreatment, 75% (6/8) of the mice from the therapy cohort receiving 1,200 μCi of ¹⁷⁷Lu-DOTA-PEG₇-Tz had tumors that were smaller than they had been on day 1.

A subjective analysis of the average tumor volume and relative tumor volume measurements (Fig. 3A and B) suggested therapeutic efficacy in both the 800 and 1,200 μCi therapy cohorts, which was sustained until the endpoint of the experiment. Some heterogeneity in response within cohorts was observed, especially for the 800-μCi cohort, as depicted by the line graphs following each individual mouse (Fig. 3C–F). This may be attributed to variations in the CA19.9 expression among tumors or differences in tumor volumes at the beginning of the study. To account for that variance and objectively examine the data over the course of the entire study, the data were analyzed using a random effect model (Supplementary Table S3). The statistical analysis showed that the therapeutic response was statistically significant for the 1,200-μCi cohort relative to both controls and the 400-μCi cohort.

⁸⁹Zr-5B1 PET images acquired in 6 selected mice at the conclusion of the in vivo PRIT study all showed demonstrable uptake of the tracer in tumor (Fig. 4A and B; Supplementary Fig. S7). The mice that received PRIT showed marginally higher uptake of the tracer in the tumor at 120-hour postinjection relative to the controls, although no conclusions as to the cause of that increase can be drawn due to the small sample size.

Ex vivo analysis of ⁸⁹Zr-5B1 distribution via autoradiography revealed a nonuniform microscopic distribution of the tracer within the tumor. The areas of highest activity concentration were observed to be close to regions with a higher density of tumor stroma (Fig. 4C), which most likely indicates delivery of the radiotracer via the tumor vasculature. Diffusion of ⁸⁹Zr-5B1 from these sites is also observed, and association with regions of 5B1 expression (Fig. 4C) is also seen. This uptake pattern of the radiotracer can be explained by a combination of slow diffusion from the site of vascular delivery and the high-affinity binding of ⁸⁹Zr-5B1 to local sites of high CA19.9 that prevents further tumor penetration. Minimal uptake of ⁸⁹Zr-5B1 was observed in regions of central tumor necrosis.

Our approach to PRIT is based on our previously reported strategy for the pretargeted PET imaging of PDAC (23). The system employed two components, a ⁶⁴Cu-labeled tetrazine radioligand (⁶⁴Cu-Tz-PEG₇-NOTA) and a trans-cyclooctene–conjugated 5B1 antibody (5B1-TCO). 5B1 is a fully human mAb that targets carbohydrate antigen 19.9 (CA19.9), a cell surface antigen and a biomarker that has shown clinical value for the diagnosis, staging, and prognostic evaluation of pancreatic cancer. Critically, our results demonstrate that the delineation of CA19.9-expressing tissues via pretargeted PET is possible despite the shedding of CA19.9 from cells and the moderate internalization rates of 5B1 after binding its antigen (23). Not surprisingly; however, a shift in the structure of the radioligand was needed. The β-emitting lanthanide ¹⁷⁷Lu³⁺ (t_1/2 ∼ 6.7 days) requires a different chelator than the divalent ⁶⁴Cu²⁺ (t_1/2 ∼ 12.7 hours) that was used for the PET imaging system. As a result, two novel tetrazine-bearing radioligands were designed for the current project, ¹⁷⁷Lu-DOTA-PEG₇-Tz and ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz (Fig. 1C). These two constructs each employ a well-characterized and clinically validated chelator for ¹⁷⁷Lu³⁺-DOTA and CHX-A″-DTPA-as well as a PEG₇ linker and Tz functionality. The inclusion of the PEG₇ linker in this type of molecule has been shown to expedite hepatobiliary clearance while maintaining the capability of similar radioligands to target TCO-conjugated biomolecules in tumor tissues.

The two radioligands that we assessed, ¹⁷⁷Lu-DOTA-PEG₇-Tz and ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz, did not produce equal tumor activity concentrations in pretargeting experiments with 5B1-TCO. Indeed, our pilot biodistribution studies in mice with BxPC3 xenografts indicated that pretargeting with ¹⁷⁷Lu-DOTA-PEG₇-Tz and 5B1-TCO produced higher tumor activity concentrations at 72-hour postinjection than pretargeting with ¹⁷⁷Lu-CHX-A″-DTPA-PEG₇-Tz and 5B1-TCO (Supplementary Fig. S3). The trends in the biodistribution data do not offer a definitive explanation for the superior performance of the DOTA-bearing radioligand compared to its CHX-A″-DTPA-modified counterpart, but the latter did show a modest increase in retention in the large intestine and kidneys (Supplementary Fig. S4). Furthermore, the stability and log D studies do little to shed any additional light on the differing uptake of the two ligands. As may have been expected, the radioligand with the cyclic chelator DOTA showed slightly higher stability in serum (Supplementary Table S1), but this would likely not account for the >2.5-fold increase in tumor activity concentrations observed in the pretargeting biodistribution experiments. Despite the lack of a cogent explanation for the differential behavior of these two radioligands,¹⁷⁷Lu-DOTA-PEG₇-Tz was chosen for more thorough in vivo biodistribution, dosimetry, and, most importantly, therapy studies.

Further biodistribution studies in which ¹⁷⁷Lu-DOTA-PEG₇-Tz was injected 72 hours after the administration of 5B1-TCO confirmed the high uptake and, perhaps more importantly, long retention of radioactivity in tumor tissue up to 5 days postinjection (Fig. 2). Clearance from nontarget tissues was rapid and led to remarkably high tumor-to-normal tissue activity concentration ratios at 120 hours (Fig. 2, inset). The gradually increasing uptake and long retention seen with this system are an advantage over other PRIT systems that rely solely on noncovalent interactions between radioligands and immunoconjugates. Ligands that rely on such interactions may clear from tumor tissue in a matter of hours after reaching maximal uptake. Of course, this is not ideal when using an isotope (e.g., ¹⁷⁷Lu) that delivers its radioactive payload over the course of weeks. By covalently linking the radioligand to the targeting vector, in this case 5B1-TCO, and achieving a rapid clearance from nontarget tissues, a higher proportion of the injected activity may be delivered to the tumor tissue while minimizing the risk of radiogenic toxicity in nontarget tissues.

The dosimetry estimations for a 70 kg Standard Man model suggest that our PRIT approach may efficiently deliver cytotoxic radiation to tumor tissues without causing significant side-effects in humans. The most common dose-limiting tissue in RIT is the red marrow, the maximum tolerated dose being approximately 150 cGy (7). Organs that are typically associated with the clearance of antibody-based agents are the kidneys, liver, and spleen, and those organs can typically tolerate doses of greater than 1,500 cGy (7, 35). The data indicate that with a single administration of 200 mCi (7,400 MBq) of radioligand, the predicted maximum tolerated dose in any off-target tissue would not be reached (Table 2). In our estimation, the MTD in humans would be reached at a dose of approximately 700 mCi (25,900 MBq), with the dose-limiting organ most likely being the red marrow (Table 2). Although it is not possible to reliably predict the radiation dose and resulting therapeutic response in human tumors, a retrospective evaluation of previous RIT approaches, both directly targeted and pretargeted, indicates that our approach meets or exceeds the dose benchmarks in preclinical murine models that would likely be necessary to achieve a therapeutic response in humans.

Previously reported PRIT systems have introduced a clearing agent via an additional injection between those of the targeting vector and the radioligand. The function of the clearing agent is to reduce the amount of accessible binding or reactive sites on the targeting vector that is still circulating in the blood at the time of radioligand injection, thus reducing the dose to sensitive, nontarget organs such as the red marrow from the subsequent injection of the radioligand. Based on our biodistribution and dosimetry results, our system may not require the use of a clearing agent, an added advantage over other systems. Previous studies have nonetheless shown that clearing agents are effective in the context of TCO- and Tz-based imaging systems (26), and this approach could certainly be adopted in the future if necessary.

PRIT studies confirmed that our approach was effective in causing significant growth delay and even regression of BxPC3 xenografts at the two highest activities tested, whereas the growth delay observed with the lowest activity (400 μCi) was not statistically significant. Although detailed statistical analysis of each cohort is complicated as mice from each cohort began to reach the endpoint, the trend is clear, and statistical significance was rapidly achieved in the control groups and both the 800- and 1,200-μCi cohorts.

Interestingly, line graphs (Fig. 3C–F) depicting the tumor size of each individual mouse within a cohort show that, perhaps with the exception of the 1,200 μCi therapy cohort, not all tumors within a particular cohort responded in a similar manner. This is likely due to differences in the initial uptake of the 5B1-TCO, the heterogeneity of CA19.9 expression, and the tumor sizes. We have observed this same variation in uptake in BxPC3 xenografts with a number of 5B1-based radiotracers in the past (23, 29, 31). In fact, heterogeneity is a common characteristic of many types of human malignancies, especially PDAC. Although no quantitative PET imaging was carried out prior to the initiation PRIT in this study, we believe that ⁸⁹Zr-5B1, which is currently being evaluated in clinical trials for patients with PDAC, may be useful as a diagnostic tool.

The results of the post-PRIT PET imaging study with ⁸⁹Zr-5B1 (Fig. 4A; Supplementary Fig. S7) suggest that CA19.9 expression persists posttherapy in tumors that remain viable after treatment. This suggests that ⁸⁹Zr-5B1 PET imaging could be used to monitor therapeutic response as well as to determine whether a patient may potentially be responsive to multiple administrations of fractionated PRIT. Pretargeted PET imaging with recently reported ¹⁸F- or ⁶⁴Cu-labeled Tz ligands would likely also be effective. Using pretargeted imaging with short-lived isotopes would rapidly provide high-quality images while reducing the radiation dose to healthy tissues. Using the pretargeted imaging approach would also provide a more straightforward assessment of the possibility of delivering additional cytotoxic radiation to the tumors via additional rounds of PRIT. Each of these potential theranostic approaches are currently being investigated in murine models of PDAC.

Ex vivo analysis showed that delivery of 5B1-based imaging and therapy agents may be limited by diffusion and trapping of the agents by areas of high CA19.9 density near sites of vascular delivery. It is likely that adjustments to the specific activity and/or injected mass of 5B1 could influence this distribution, as could the length of time allowed between administration and assessment. PDAC tumors are known for their dense desmoplastic stroma, and tumor penetration may well be a concern in the translation of antibody-based PRIT. However, this constraint may be alleviated by the use of compounds that act to increase stromal permeability (36). Further studies to evaluate the effectiveness of such an approach are currently underway in our laboratory.

In conclusion, our in vivo therapy results have shown that our bioorthogonal pretargeting system utilizing Tz-based radioligands and TCO-modified immunoconjugates is a promising platform for the development of PRIT systems. It is important to note that the inherent complexity of pretargeted immunotherapy can create complications as we move toward the clinic. Indeed, the translation and clinical optimization of a multicomponent system will undoubtedly be more cumbersome than analogous approaches that utilize a single agent. Nonetheless, we are confident that the benefits of our PRIT strategy, and especially future generations, will outweigh any logistical drawbacks. Along these lines, we believe that this platform warrants further optimization and a number of prospective routes could lead to improved results. For example, previous studies on PRIT systems that use a fractionated dosing schedule have shown promise, with some demonstrating equal absorbed doses in nontarget organs and significant improvements in response rates compared to a single administration of the radioligand. Our laboratories are currently working to establish MTD in preparation for fractioned PRIT studies using the system reported herein. In addition, in future experiments we will assess the ability of 5B1-based PET imaging to quantify CA19.9 levels prior to therapy in order to evaluate ⁸⁹Zr-5B1 as a theranostic tool for predicting response to 5B1-based PRIT.

No potential conflicts of interest were disclosed.

Conception and design: J.L. Houghton, W.W. Scholz, J.S. Lewis, B.M. Zeglis

Development of methodology: J.L. Houghton, S. Carlin, J.S. Lewis, B.M. Zeglis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.L. Houghton, R. Membreno, D. Abdel-Atti, S. Carlin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.L. Houghton, R. Membreno, K.M. Cunanan, S. Carlin, W.W. Scholz, P.B. Zanzonico, J.S. Lewis

Writing, review, and/or revision of the manuscript: J.L. Houghton, R. Membreno, D. Abdel-Atti, W.W. Scholz, P.B. Zanzonico, J.S. Lewis, B.M. Zeglis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Membreno, D. Abdel-Atti, J.S. Lewis

Study supervision: J.S. Lewis, B.M. Zeglis

We would like to thank the staff of the Small Animal Imaging Core Facility, the Radiochemistry and Molecular Imaging Probe Core Facility, Mr. William H. and Alice Goodwin and the Commonwealth Foundation for Cancer Research, and The Center for Experimental Therapeutics at MSKCC.

The MSKCC Small Animal Imaging Core Facility as well as the Radiochemistry and Molecular Imaging Probe core were supported in part by NIH grant P30 CA08748. J.L. Houghton received 1F32CA180452-01A1 and 5R25CA096945-09. W.W. Scholz received 2R42CA128362 and HHSN261201300060C. B.M. Zeglis received 1K99CA178205-01A1 and 4R00CA178205-02. All experiments were supported by The Experimental Therapeutics Center at MSKCC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.